Murphy's ABC protocols Consider, as an illustration, the predicament of Murphy, a clinical supply specialist. His protocols, A, B, and C all use
a dispensing unit type called "10mg Samplovir." In this example, we will focus on only one physical location, "Fenway," where
all three protocols are being conducted. Doing what he's always done, Murphy packages "Lot 1" of 10mg Samplovir, labels, electronically marks it in the database for
protocols A, B, and C, and distributes it to his supply chain. In marking for specific protocols, Murphy has broken pooling's fundamental principle that dispensing units must be protocol
independent. But the protocols for which he has labeled happen to be the only protocols that use 10mg Samplovir, so all drug
need calculations—for plotted subject consumption, safety stock, floor/ceiling—are successfully totaled. Now, however, Murphy packages a "Lot 2" of 10mg Samplovir. He feels that protocol C is just about wrapped up and does not
want to dedicate any new supply to it. So this time he only marks it for protocols A and B. Murphy distributes Lot 2, and
some ends up at Fenway alongside Lot 1. Murphy's supply at Fenway is now comprised of 17 total dispensing units of 10mg Samplovir: two units of 10mg Samplovir from
Lot 1 and 15 units of 10mg Samplovir from Lot 2. Lot 1 is usable for protocols A, B, and C, but Lot 2 is only usable for protocols
A and B. In this example, he still has subjects enrolled and active at this location for all three protocols. That night, Murphy's resupply algorithm now begins counting his supply at Fenway. When it counts drugs at location, it can
no longer count all 17 units of 10mg Samplovir together in a single bucket. At a minimum, it needs to count once for protocols
A, B, and C (two units) and again for protocols A and B together (15 units). There is no way for the algorithm to represent
the stock with a single number for the location. At the moment, however, this is little more than an inconvenience. The following week, Murphy needs to supply drugs for projected subject need. Protocols A, B, and C have two subjects each
scheduled to arrive in the next week; each subject is scheduled to receive one 10mg Samplovir. Ordinarily, a pooled supply
algorithm would compare this summed need (six units) against stock at the site (17 units), but since Murphy broke the absolute
nature of the pool, the algorithm must count the needs against the stock at site separately. If the algorithm does this for each protocol separately, it may predict that the needs of protocols A and B for two each will
be easily supplied by their total available stock of 17 and 17, respectively, and that protocol C's need for two will also
be covered by the remaining Lot 1 supply of two. But the algorithm's prediction may not necessarily be correct. 
Scheduled Dispensing Visits at Fenway
|
Consider the drug consumption for randomized subjects scheduled to have dispensing visits at Fenway during the following week.
If both subjects from protocol C arrive earlier than subjects from protocols A and B, one should see something like what appears
in Table 1.
Fenway Encounters Dispensing Problems
|
The order of the last four subjects, in this case, does not really matter; it matters only that the subjects from protocols
A and B all arrive after the subjects from protocol C. But what happens if a subject from protocols A or B arrives prior to
a subject from protocol C? The dispensing algorithm should be dispensing the oldest usable medication first so Lot 1 will
be consumed by subjects in protocols A and/or B (see Table 2).
Algorithm vs. IVRS The more subjects are intermingled at a location and the more stock is labeled for different subsets of protocols, the worse
the problem gets. Even in this example, if subjects from A or B arrive in the first two dispensing positions, both subjects
from protocol C will encounter stock-outs and be lost. Losing subjects in this manner is clearly unacceptable. Why can't an
algorithm figure out on the fly which lots are for which sets of protocols and do the math accordingly? While it may seem intuitive that the algorithm should be able to do modified calculations to rescue Murphy, or at least to
set aside the Lot 1 stock for the protocol C subjects, in fact the algorithm is ill suited to the task. To preserve certain
drugs for the projected need of specific incoming subjects, the dispensing algorithm would need to be linked to the predictive
resupply algorithm, and the two functions are mismatched. Dispensing in IVRS studies is an uncomplicated operation: The algorithm looks at the dispensing unit type(s) required by the
subject, chooses the earliest expiring usable unit(s) from stock at the site, and dispenses. This simple process is carried
out in real time many times a day to provide drugs to subjects. A resupply algorithm, by contrast, is a very complex process
that usually runs once a day and considers numerous factors and projections in its decision making. Conclusion All in all, drug pooling has the potential to increase supply efficiency in an unexpectedly wide range of scenarios, although
its limits may be unexpected. As the inherent flexibilities and inflexibilities of the technique—and the breadth of situations
in which it can help contain costs—become more commonly understood, the use of drug pooling will undoubtedly continue to expand.
Dave Riege is associate research fellow in supply chain management at Pfizer. Ed Tourtellotte* is the founder and chief executive officer of Tourtellotte Solutions, 321 Commonwealth Road, Suite 303, Wayland, MA 01778,
email: etourtel@tci9.com
*To whom all correspondence should be addressed. References 1. L. George, "Investigational Medicinal Products—Optimizing the Supply Chain," Applied Clinical Trials, April 2005, 42–48.
|
